On Sun, 16 Jun 2002, Makoto Fujiwara wrote:

> I don't have enough time today, sorry.

That's ok.

> Two sets of patches Bill provided doesn't give me any changes
> when I have applied them to src on 20020612 or so.
> (I am not really sure on this point, I need time to make sure)

The second patch (third diff), the one that adds the call to
cpu_probe_cache(), didn't fix things?

Please make sure of this.

> The situation I have: (with kbd/screen)
>
> (1) boot.fs snapshot/20020407-1.5ZC   ...  stops at start=0x100000
> (2) kernel with -D20020407-UTC syssrc ...  stops at start=0x100000
> (3) kernel with 20020314 + 20020313   ...  boot fine
>
> All three boot fine if ttya is used for the console.
>
> So what we have to look at is diff between
> 20020314+0313 and 20020407 ?

No, we don't. The purpose of the timewarp is to have
lib/libkern/arch/powerpc at 20020313 and the rest of the kernel at
20020314. Manuel (and I thought you also) found that a kernel from
20020314 doesn't boot at all. So if reverting lib/libkern/arch/powerpc
fixes things (as the time warp indicates), then the change that causes the
problem is the one to lib/libkern/arch/powerpc. There are two files that
were changed at that date, but only one is actually used.

So that means we should have a LOT less code to need to look at.

That change made __syncicache() do nothing if the cpu hasn't been attached
as part of autoconfig yet. There are good reasons for doing this (which
have to get into ppc specifics), but it's not what macppc expects.

Take care,

Bill